Identification and comparative analysis of potholes using image processing techniques

Abstract

Potholes have become major havoc and are the leading reason for the damage of road transport vehicles. Hence, it is important to asses this problem and to provide a solution which can aid the driver of the vehicle before approaching a pothole. The topic selected uses 4 different image segmentation techniques to identify all types of potholes. The following techniques that were worked on were Image Thresholding, Canny Edge Detection, K-Means clustering, and Fuzzy C-Means clustering. Performance analysis of the different image segmentation techniques is done in MATLAB 2015Ra image processing toolbox. The effectiveness of the different image segmentation techniques was then tested in different environments. Thus the results were generated in terms of accuracy and precision. Moreover, the results were compared with each other to draw a conclusion on their viability. Finally, the paper emphasizes why these techniques are good for developed infrastructure and why they are not for third world countries.